Fig. 1.

(A) Relative efficiency of unweighted average compared with the estimate from MLM. If it is close to 1, standard errors of unweighted average and MLM estimate are similar. The less it is, the less standard error MLM has compared with the unweighted average. The technical variation is fixed constant across hundred thousand samples at σ and the biological variation τ are set to be 0, 10, 25, 50 or 75 of σ. The bead numbers were first generated from normal distribution with mean 40, and then rounded and replaced with five if the number is less than five. We repeated the process of computing the relative efficiency by gradually increasing the variation of the bead numbers. (B) Relative efficiency of the WLS^a estimate under our proposed model MLM.